Scientists give talks at conferences, local seminars, or workshops. Some of them are recorded and below are a few for which there is an online version, as well as a couple for which I only have the slides.
This first talk is about what the deal.II library is and where it is located in the open source world. I gave this talk at the Free and Open Software Development Meeting 2020 (FOSDEM'20) conference in Brussels, in February 2020 – just before the global Coronavirus stuff happened. It was a packed conference, with many rooms standing room only and restricting more people from coming in. It was a fun environment and, probably, unbeknownst to most of us the last conference or meeting we would go to for quite a while.
For this talk, I don't actually have an embeddable video – but you can find it here.
This is a colloquium talk in the Colorado School of Mines's Applied Mathematics and Statistics department. It covers the motivation and details for simulating convection in the Earth mantle. As an added bonus, the first ten or so minutes covers an interview by Paul Constantine in which he asked me a number of questions about my work and other things.
A talk that I gave as part of a Webinar series for the Computational Infrastructure in Geodynamics initiative, an NSF funded center dedicated to providing the geophysics community with high-quality computational software. The talk is about what can be achieved by using existing software libraries.
The youtube video above shows the slides and has my voice as audio. (During processing, the audio was lost between minutes 1 and 3 but resumes after that.)
This is a talk I gave while visiting my friend and colleague Moritz Diehl, an expert in optimization theory, in Leuven, Belgium. It is a talk about the many steps one has to go for biomedical imaging, but actually starts explaining the reasons why one would need yet another one of these imaging methods. It goes all the way from introducing a simple model problem to inverting real measurement data with an adaptive finite element program.
A talk given in January 2005 in Pittsburgh. The topic is duality-based error estimates. I give an introduction into why the estimation techniques usually used today are not really satisfactory in practice and how duality-based estimates can mitigate this. I then move on to show how such techniques can be applied to inverse problem, a case where traditional estimation techniques have completely failed because the problems are not only nonlinear but also ill-posed and the stability estimates one usually needs are unknown or inexistant.